Mass production was originally slated for 2028<\/a>, but Kuo now says it could start as early as the first half of 2027, likely tied to OpenAI’s planned IPO<\/a> and the heating-up AI hardware race.<\/p>\n But OpenAI’s pitch is the software running on the hardware: an AI agent that handles tasks for users instead of routing them through separate apps, a service that might work better if OpenAI controls the hardware and OS itself. The company has the brand, the user data, and the models to try it, with revenue likely coming from a mix of subscriptions and hardware sales.<\/p>\n MediaTek has the upper hand as the sole chip supplier, and combined shipments for 2027 and 2028 could hit around 30 million devices, according to Kuo.<\/p>\n Picking a smartphone as the form factor is also\u00a0a quiet admission that more experimental AI hardware isn’t ready for prime time yet<\/a>. Smartphones still dominate by a wide margin, and Google already has a massive lead through Android, with its own agentic AI features in the works<\/a>.<\/p>\n The biggest issues with these systems are still security and reliability, and even if the software worked perfectly, there’s no killer use case yet beyond “nice to have” tasks like booking a restaurant. So these plans have mild “Facebook Phone” vibes<\/a>, which was also just an OS reskin slapped on existing hardware. But who knows, OpenAI might knock it out of the park.<\/p>\n \t\t\t\tAI News Without the Hype \u2013 Curated by Humans<\/p>\n \n\t\t\t\t\tSubscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive “AI Radar” frontier report six times a year, full archive access, and access to our comment section.\t\t\t\t<\/p>\n![\"\"](\"https:\/\/www.europesays.com\/ai\/wp-content\/uploads\/2026\/05\/kuo_agent_phone_vision.png\")
A conceptual visualization of OpenAI’s agent software by analyst Ming-Chi Kuo shows how today’s app icon grid could evolve into an agent task stream. | Image: Ming-Chi Kuo<\/a><\/p>\n